Can regional resource shares be based only on prevalence data? An empirical investigation of the proportionality assumption.

The needs component of the current formulae for allocating resources for hospital services and prescribing in England is based on a utilisation approach. This assumes that expenditure on NHS activity in different geographical areas reflects relative needs and supply conditions, and that these can be disentangled by regression models to yield an estimate of relative need. These assumptions have been challenged on the grounds that the needs of some groups may be systematically 'unmet'. Critics have suggested an alternative based on variations in the prevalence of health conditions, called the 'epidemiological approach'. The epidemiological approach uses direct measures of morbidity to allocate health care resources. It divides the total national budget into disease programmes based on primary diagnosis, computes the proportion of total cases for each programme in each geographical area, and then allocates budgets to geographical areas proportional to their share of total cases. The main obstacle to the epidemiological approach has been seen as its very demanding data requirements. But it also faces methodological challenges. These centre on the assumption of proportionality which, at the area level to which resources will be allocated, requires that the average level of need for 'cases' within each disease programme is the same in every area. We illustrate the epidemiological approach, and test the proportionality assumption underpinning it, using data from the 2002-2004 rounds of the Health Survey for England. We find regional variation in disease severity for major diseases, which suggests that health care needs for some conditions vary by area. Further analysis suggests that the epidemiological approach might systematically underallocate resources to rural areas, areas with younger populations, and deprived areas. Since the proportionality assumption underpinning the epidemiological approach does not hold, its adoption would fail to take account of variations in severity. This casts some doubt on the utility of the approach for resource allocation at the present time.